Playwright HTTP Cache & Revalidation Auditor

Compare a cold browser visit with a repeat visit to a public page. Get cache-hit evidence, 304 revalidation, summarized cache policy, and transfer/timing deltas without storing payloads or raw URLs.

Example

{"urls":["https://example.com/"],"settleMs":500,"navigationTimeoutSeconds":20,"maxResources":100}

Supply 1-3 public URLs. Each URL receives its own fresh browser context, two ordinary navigations, and one report. Inputs are deduplicated. settleMs is 0-3000; navigationTimeoutSeconds 5-30; maxResources 1-200 per visit. No credentials, cookies, headers, custom scripts, proxy configuration or interactions are accepted.

Use cases

• Verify browser cache reuse after CDN/cache-header changes.
• Inspect validators and 304 behavior on deployment assets.
• Compare cold/repeat transfer as a focused complement to HAR diagnostics.
• Add repeatable browser cache checks to release QA.

Output

Dataset AUDITED rows contain cold and repeat visit summaries and bounded resource records. Each phase has durationMs, resourceCount, cacheHitCount, revalidatedCount, networkBytes, failedRequests and truncated. Resource records have hashed identity, origin, resource type, logical status, wireStatus, cache source, CDP encoded bytes and policy flags. cacheSource browser is an unspecified browser-cache signal, not a guaranteed memory-cache classification. wireStatus may be null if no wire response exists.

transferDeltaBytes = cold minus repeat CDP byte totals; durationDeltaMs = cold minus repeat elapsed times. Positive values do not prove a general speed improvement. OUTPUT summarizes outcomes. cache-* KVS reports duplicate the structured evidence for integration.

Safety and limitations

Private, loopback, metadata, reserved and mixed DNS destinations are blocked by a DNS-pinned proxy. Service workers and downloads are disabled. Do not interpret serviceWorkerMode blocked as evidence that the target has no service worker. This does not test PWA offline behavior or cache invalidation across deployments. No browser routing is used because it disables cache.

Page and resource paths/query values are hashed; only origins remain readable. No response bodies, cookies, header contents or validator strings are persisted. Policy fields are flags, safe numeric durations and validator-presence booleans. Browser background/late traffic, resource caps, cookies created during the first visit, and dynamic pages can affect comparison. HTTP cache coldness is local to the fresh browser; CDN/origin caches may already be warm. Byte totals are CDP evidence, not billing-transfer measurements, origin hit counts, or Lighthouse/Web Vitals scores.

Pricing proposal

Standard tier: USD 0.012 per cache-pair-audited event after dataset and KVS evidence persist. Failed pairs free. A successful pair with no cache hits remains a useful audit. Platform costs depend on plan and usage. The Actor is configured for pay per event at USD 0.012 per successfully stored pair.

Suggested tasks

Cold/repeat homepage check; deployment asset revalidation audit; compare public landing-page cache behavior. Repeat over time under consistent browser/settings; no automated load testing.

Local execution

Node 22+, locked npm dependencies, Chromium executable in APIFY_CHROME_EXECUTABLE_PATH. npm test runs meaningful browser fixtures. Put input at storage/key_value_stores/default/INPUT.json and npm start. Tests have an internal loopback fixture option that production input never exposes.